DESCRIPTION (Taken from application abstract): In the rapidly changing health care environment the increasing prominence of managed care has prompted a greater reliance on formal clinical guidelines to suppose clinical decisions. Guidelines have been advocated with increasing frequency to reduce inappropriate care, control geographic variations in practice patterns and make more effective use of health care resources. However, guidelines often have little impact on clinical practice because physicians are unaware of them, lack confidence in them because the justification for their recommendations is not clear or because they are inaccessible at the time of patient care or difficult to apply. Guidelines also quickly become out of date as new research data becomes available. In order to enhance the quality and usefulness of clinical guidelines, the general goal of this project is to develop, deploy and evaluate interactive computer based guidelines, supported by an integrated decision theoretic model and a linked knowledge base. This arrangement will use patient characteristics to tailor guideline advice to the individual patient. The integrated decision model will provide recommendations for situations not addressed by the guideline and also will help to justify guideline recommendations by calculating the effectiveness and costs of various strategies. The proposed system also contains links to a knowledge base containing the sources of data used in the guideline and the model, so that physicians using the system can examine the studies supporting the guideline recommendations. Natural language explanations, generated automatically based on the structure of the decision model and the guideline will justify the recommendations. A query capability will enable physicians to look up specific data from the knowledge base. The linked knowledge based also ensures that the supported decision model and guideline will be updated automatically as new research data is published. The computer system will be used to implement an interactive version of the Guidelines for Medical Treatment for Stroke Prevention, developed by the American College of Physicians. The system will be bench tested using a series of cases abstracted from the General Medicine practice at the Robert Wood Johnson Medical School. Faculty and house staff internists will serve as research subjects to perform a field trial of the system. This will include the extent of previous compliance with the guideline, pre and post-testing of medical knowledge pertinent to the guideline, and the degree to which the computer-based guideline changes behavior compared to the traditional guideline format. A decision theoretic measure of potential benefit will be calculated by comparing decision model evaluations of physicians' unaided choices with those recommended by the guideline system.